Objections

After this lesson, the reader should understand:

• 01State each recurring objection in its strongest form.
• 02Give the substantive answer, not a rebuttal by assertion.
• 03Identify the objection that is partly correct and what it correctly identifies.

The strongest objection

'What if the institution encodes its authority incorrectly? You have just given a wrong rule a runtime and a seal.' What is the honest answer?

That the risk is real and the discipline does not eliminate it^[1]. What it does is make the failure visible instead of silent. An incorrect prose policy fails through inconsistent human application that no one records. An incorrect artifact fails the same way a defective program does — reproducibly, against a version, traceable to the compilation step that introduced it — and the response is the same: source provenance, review by the authority holder, conformance testing, version control, independent inspection^[2]. The objective was never a representation that cannot be wrong. It is that an error in authority does not get to become invisible software behaviour. The five objections share a blind spot: each accepts that the right party is doing the encoding. A sixth does not, and it may be the most serious. Who authorised the compilers? A platform that compiles public law into sealed artifacts and publishes them for anyone to query is offering interpretations, and interpretations that are free, fast, and machine-callable have a way of becoming authority by convenience. If a thousand agents ask a published reading of a regulation whether an action is allowed, that reading governs a thousand actions — whatever the legislature meant, and whether or not the regulator ever endorsed it. Scholars of legal technology have warned for years that the design choices of code can displace legal judgement without anyone deciding they should^[3]. This objection is partly correct, and the part that is correct changes the design rather than just its framing. The answer cannot be that third parties must not publish interpretations; lawyers, textbooks, and compliance vendors always have. It is that an artifact must say whose interpretation it is, and a caller must be able to tell the difference. A reference artifact should declare itself as reference — the publisher's reading, not the issuer's answer — in a field that travels with every verdict, so that an agent relying on it, and anyone auditing that agent, sees the standing as clearly as the result. The worked examples in this curriculum carry exactly that disclosure. Whether disclosure is enough to stop convenience from turning into authority is an open question, and it is one the field should keep asking. A seventh objection arrived from a reader of this curriculum^[4], and it is worth stating because it sharpens the other six. A verdict is only as strong as its force: a verifier that sits inside the agent produces advice, because the agent can ignore it, and a correct 'no' that changes nothing is a log entry, not a decision. The objection is right, and the discipline's answer is architectural. The verdict must be computed by the system whose state the action would change, and that system must be the only path to the change. Every negative verdict then has a destination: incomplete routes to evidence, escalated routes to the authority the rules designate, and denied ends the autonomous path. The reader's challenge also prompted a re-test of Lecture 4.4's own case study against the specification, and the re-test found a separate gap: a superseded but unrevoked version of an authority could still authorise a new action^[5]. It is now closed, and the change records what prompted it. That sequence — a claim, an outside challenge, a re-test, a defect, a specification change, a test that pins it — is the strongest available answer to the objections in this lecture, because it shows the discipline can be wrong in public and correct itself in a way anyone can check.

• Which of the five objections is strongest, and does it change the design or just its framing?
• Is 'partly access control' a fair description, and where does the analogy break?
• What review regime would you require before an artifact could execute?
• Does making interpretation explicit reduce discretion, or relocate it?
• If a third party's published reading of a regulation is what most agents actually consult, whose authority governs their actions?
• Is a 'reference' label on an artifact enough to keep convenience from turning into authority?

No responses yet. Be the first to answer one of the questions above.

Steelman one objection and answer it.

1. 01Pick the objection you find most persuasive.
2. 02Write it in its strongest form, with a concrete example.
3. 03Write the answer, conceding whatever is true in it.
4. 04State what would have to change in the architecture if the objection held.